Re: Simulating different populations

From: "Gastonguay, Marc" marcg@metrumrg.com Subject: Re: [NMusers] Simulating different populations Date: Thu, 11 Aug 2005 08:59:42 -0400 Hi, Pascal. Thanks for pointing out these limitations and assumptions. I think your point about fixed effects parameters that have natural bounds is an excellent one; we do we have to think carefully about the parameterization and choice of posterior distributions. For this reason it may be simpler (but more computationally intensive) to simulate from empirical distributions, such as those that might result from a nonparametric bootstrap. For the implementation part, I agree that an automated extraction of relevant components is mandatory. It sure would be nice to have an all-in-one tool that did this for us, but I think that this discussion also illustrates the wonderful flexibility of an ASCII text-based, command-line program like NONMEM, when coupled with a useful programming language (we use R and Perl). Thanks for the helpful suggestions. Marc _______________________________________________________